| 基于音素分类的汉语语声转换算法 |
| |
| 引用本文： | 李燕萍,张玲华,丁辉. 基于音素分类的汉语语声转换算法[J]. 南京邮电大学学报(自然科学版), 2011, 31(1): 10-15,38 |
| |
| 作者姓名： | 李燕萍 张玲华 丁辉 |
| |
| 作者单位： | 1. 南京邮电大学,通信与信息工程学院,江苏,南京,210003
2. 嘉兴学院,数学与信息工程学院,浙江,嘉兴,314001 |
| |
| 基金项目： | 国家自然科学基金，南京邮电大学引进人才基金，浙江省自然科学基金 |
| |
| 摘 要： | 提出了一种基于音素分类的语声转换算法。根据汉语语音学理论,分析汉语语言结构,通过帧特征分析,构建了汉语元音映射表,指导音素的分类,然后对每类音素进行独立的特征参数转换,解决了传统GMM转换算法的过平滑问题,避免了训练与测试阶段的语料音素之间的不均衡,从而提高语音的可懂度与自然度。客观评测结果表明,基于音素分类的语声转换算法与GMM算法相比,谱距离比值平均降低了9.31%;主观听力测试表明,ABX测试结果提高了10.36%,同时MOS得分平均提高了29.33%。
|
| 关 键 词： | 语声转换 音素分类 汉语元音映射 高斯混合模型 |
A Novel Algorithm for Chinese Voice Conversion Based on Phoneme Classification |
| |
| LI Yan-ping,ZHANG Ling-hua,DING Hui. A Novel Algorithm for Chinese Voice Conversion Based on Phoneme Classification[J]. JJournal of Nanjing University of Posts and Telecommunications, 2011, 31(1): 10-15,38 |
| |
| Authors: | LI Yan-ping ZHANG Ling-hua DING Hui |
| |
| Affiliation: | LI Yan-ping1,ZHANG Ling-hua1,DING Hui2 1.College of Telecommunications & Information Engineering,Nanjing University of Posts and Telecommunications,Nanjing 210003,China 2.College of Mathematics and Information Engineering,Jiaxing University,Jiaxing 314001,China |
| |
| Abstract: | A novel algorithm for Chinese voice conversion based on phoneme classification is proposed in this paper.The analysis of Chinese linguistic structure provides the construction of vowel mapping theory which guides the classification of phonemes.The proposed method implements feature parameters conversion for each category phoneme independently.It deals with the spectral smoothing of traditional GMM(Gaussian Mixture Model) and prevents phoneme imbalance between training and testing materials and then improves... |
| |
| Keywords: | voice conversion phoneme classification chinese vowel mapping gaussian mixture model |
| 本文献已被 CNKI 万方数据 等数据库收录！ |
|
